Peter T. Leeson and Jacob W. Russ from George Mason University collected this data to explain the surge in witch trial activity in 1550–1700. They claim it happened because of escalating competition between Catholic and Protestant churches. Other popular theories are that: bad weather, i.e., so-called "witches," became scapegoats for crop failures and famine caused by climate change; and a lack of governance.
The trials dataset contains information about almost 7,000 witch trials:
The battles dataset includes information about European religious wars and battles of the time, collected by the authors to prove their theory about confessional rivalry.
I vizualise the data here, and in this notebook I explain how it was processed and analysed. To put the data on a map, I brought it in line with EU NUTS levels.
Trials dataset:
| year | decade | century | tried | deaths | city | gadm.adm2 | gadm.adm1 | gadm.adm0 | lon | lat | record.source | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | NaN | 1520 | 1500 | 1 | 1.0 | NaN | NaN | NaN | Estonia | NaN | NaN | Madar (1990) |
| 1 | NaN | 1530 | 1500 | 1 | 1.0 | NaN | NaN | NaN | Estonia | NaN | NaN | Madar (1990) |
| 2 | NaN | 1540 | 1500 | 5 | 5.0 | NaN | NaN | NaN | Estonia | NaN | NaN | Madar (1990) |
| 3 | NaN | 1580 | 1500 | 7 | 5.0 | NaN | NaN | NaN | Estonia | NaN | NaN | Madar (1990) |
| 4 | NaN | 1590 | 1500 | 11 | 0.0 | NaN | NaN | NaN | Estonia | NaN | NaN | Madar (1990) |
1. The exploration will be based on decades.
There are three time intervals in the dataset:
But the "Year" column has 9% of NaN values, while the "Decade" has none, so I'll use the latter.
2. To be put on a map with the help of EU GeoJSON data, the regional structure of the dataset needs to be transformed.
In the original dataset, the territory is divided by:
In fact, the dataset has been collected from different sources, so regions and counties are not always comparable. In some countries, the territory listed in the "region" column corresponds to the NUTS-3 level of EU statistical division; in others, it's NUTS-1, etc. Moreover, some regions and counties in the dataset no longer exist in official statistics. There are "longitude" and "latitude" columns, but unfortunately, they don't cover even half of the data, so it's better not to rely on them. That's why I remap the witch trials according to EU NUTS levels as of 2021.
Battles dataset:
| battle | month | year | decade | century | city | country | war | war.theater | lon | lat | gadm.adm0 | gadm.adm1 | gadm.adm2 | source | source.page | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | Trier | August | 1522 | 1520 | 1500 | Trier | Germany | Knights' Revolt | NaN | 6.637143 | 49.749992 | Germany | Rheinland-Pfalz | Trier | Hitchcock (1958) | 1 |
| 1 | Landstuhl | April | 1523 | 1520 | 1500 | Landstuhl | Germany | Knights' Revolt | NaN | 7.570905 | 49.414146 | Germany | Rheinland-Pfalz | Rheinhessen-Pfalz | Hitchcock (1958) | 1 |
| 2 | Donaueschingen | December | 1524 | 1520 | 1500 | Donaueschingen | Germany | German Peasants' War | NaN | 8.497946 | 47.952152 | Germany | Baden-Wurttemberg | Freiburg | Miller (2003) | 17 |
| 3 | Kempten | April | 1525 | 1520 | 1500 | Kempten | Germany | German Peasants' War | NaN | 10.315784 | 47.728570 | Germany | Bayern | Schwaben | Miller (2003) | 14 |
| 4 | Leipheim | April | 1525 | 1520 | 1500 | Leipheim | Germany | German Peasants' War | NaN | 10.226306 | 48.444412 | Germany | Bayern | Schwaben | Miller (2003) | 14 |
There are no NaN values in the battles dataset, so I'll just use the longitude and latitude columns.
New regional division by NUTS (0 = country; 1-3 = corresponding NUTS levels):
| country | nuts_level | |
|---|---|---|
| 0 | Hungary | 0.0 |
| 1 | Finland | 1.0 |
| 2 | Norway | 1.0 |
| 3 | Estonia | 2.0 |
| 4 | Poland | 2.0 |
| 5 | Italy | 2.0 |
| 6 | Netherlands | 2.0 |
| 7 | Spain | 2.0 |
| 8 | Luxembourg | 2.0 |
| 9 | United Kingdom | 2.0 |
| 10 | Ireland | 2.0 |
| 11 | Belgium | 2.0 |
| 12 | France | 2.0 |
| 13 | Austria | 2.0 |
| 14 | Germany | 2.0 |
| 15 | Denmark | 2.0 |
| 16 | Switzerland | 3.0 |
| 17 | Czech Republic | 3.0 |
| 18 | Sweden | 3.0 |
How much data is lost after remapping?
| Country | Total_Rows | NaN_OLD_region | NaN_NEW_region | NaN_OLD_region_% | NaN_NEW_region_% | |
|---|---|---|---|---|---|---|
| 0 | Estonia | 17 | 17 | 0 | 100.0 | 0.0 |
| 1 | Finland | 17 | 17 | 0 | 100.0 | 0.0 |
| 2 | Hungary | 26 | 26 | 0 | 100.0 | 0.0 |
| 3 | Norway | 20 | 20 | 0 | 100.0 | 0.0 |
| 4 | Switzerland | 862 | 6 | 6 | 0.7 | 0.7 |
| 5 | Germany | 2904 | 4 | 59 | 0.1 | 2.0 |
| 6 | Austria | 12 | 4 | 4 | 33.3 | 33.3 |
| 7 | France | 614 | 51 | 51 | 8.3 | 8.3 |
| 8 | Belgium | 494 | 0 | 0 | 0.0 | 0.0 |
| 9 | Ireland | 4 | 0 | 0 | 0.0 | 0.0 |
| 10 | United Kingdom | 932 | 0 | 135 | 0.0 | 14.5 |
| 11 | Luxembourg | 20 | 1 | 0 | 5.0 | 0.0 |
| 12 | Spain | 29 | 0 | 0 | 0.0 | 0.0 |
| 13 | Netherlands | 300 | 4 | 4 | 1.3 | 1.3 |
| 14 | Italy | 103 | 1 | 1 | 1.0 | 1.0 |
| 15 | Poland | 9 | 0 | 0 | 0.0 | 0.0 |
| 16 | Czech Republic | 1 | 0 | 0 | 0.0 | 0.0 |
| 17 | Sweden | 276 | 0 | 0 | 0.0 | 0.0 |
| 18 | Denmark | 51 | 6 | 6 | 11.8 | 11.8 |
Only the UK has significant losses in values after remapping, compared with the initial division by the 'gadm.adm1' column; however, for the U.K., this column is detailed to the level of England, Scotland, etc., which is too broad for meaningful regional exploration and mapping.
The dataset contains information about 39,052 people prosecuted for witchcraft between 1300 and 1850. 15,874 of them were sentenced to death (41%). The real numbers are estimated at 100,000–110,000 of accused and 40,000–60,000 of executed.
I didn't find information about some famous trials in the dataset, such as Bamberg (1,000 killed) and Fulda (250 killed), perhaps just because of its detalization.
Before ~1550, witch trials seem to have been sporadic. Then their number started to grow fast, and peaked in the 1610-1619. The majority of trials didn't end in killings in that decade, however; deadly cases reached their peak a decade later. All in all, the "active phase," if we can call it that, lasted for about 150 years; it took another 150 years for witch trials to disappear completely.
The territory of witch trials and executions seems to be unified and compact.
85% of all witch trials took place in Germany, Switzerland, France, Spain, and the UK. 94% of all executions were held in Germany, Switzerland, France, Hungary, and the UK. The citizens with the fewest chances to leave after the trial were those of Switzerland, Luxembourg, Germany, France, and Belgium (though we should remember that, besides being executed, people could die in custody, as happened to many victims of the brutal Würzburg witch trials).
Three groups of countries stand out here:
Leeson and Russ claim that the unpopularity of witch hunts in some countries may have been connected to the absence of competition between the Catholic and Protestant churches. For example, in Spain, where Reformation ideas didn't find many followers, witch trials didn't become widespread. They were more popular in the territories where religions became rivals.
Big number of accused for witchcraft not necessarily ended in large scale of executions. For example, the biggest part of trials took place in Germany's Mecklenburg-Vorpommern, and there's no evidence of any executions as a result of those trials; however, in the Swiss canton of Vaud, which was the third in terms of the number of people tried, mortality reached 100%, which accounts for 21% of all European executions. Same with the cities. Besides the scale of religious controversies inside a particular area, the ruler's personality could play a role:
"Among prince-bishops, Philipp Adolf von Ehrenberg of Würzburg was particularly active: in his reign of eight years (1623–31) he burnt 900 persons, including his own nephew, nineteen Catholic priests, and children of seven who were said to have had intercourse with demons." Hugh Trevor-Roper, The Crisis of the Seventeenth Century
Depending on the country, the witch hunt was predominantly an urban or rural phenomenon: in Germany, Belgium, and Luxembourg, the majority of witch trials and executions took place in cities, while in Switzerland they occurred mostly in the country.
Well, it seems like, on the regional level, more trials occured in places where the battles were not held. Seems that they mach in time, not in place. Let's check whether there's a dependency.
I took the period of European religious wars (1520–1650) and tried to find a correlation between battle and trial activity. Well, there seems to be a moderate effect on the country level; however, I failed to find any on the regional level. On the contrary, it seems that witch trials mostly took place in peaceful regions.
Checking the map, in my opinion, supports the idea that while witch trials indeed surged at the time of religious wars, they mostly took place on the territories free of battles.
So, could there be any substitutional effects?..